On 01/31/2011 06:08 PM, Paul Eggert wrote: > Thanks for reporting the bug, but I'm afraid that we can't > easily reproduce it from the info in that message. Also, > it's not clear what patches have been applied to the Solaris > distribution of GNU tar. So you may have to provide more > debugging information.
May i also point you to [1]. I'm experiencing timestamp related problems, which may be related too. There is a patch too, which i wanted to submit for review anyway. The strange thing about this bug is the hard-to-reproducibility. Looks like a heisenbug (in gnulib?). When i set a breakpoint after setting the timestamp, and the program stops, the timespamp of the extracted file is correct, but immediately overwritten again. If i set a breakpoint _before_ setting the timestamp, correctness of the end result depends on how long i stay at this breakpoint. it seems if i immediately continue, the timestamp is set correctly but immediately overwritten again. however if i spend a second or so at the breakpoint, the timestamp is not only set correctly, but also persists the close() call, which isn't the case otherwise. All this observations where done on i586-pc-interix6.1, so i have no idea whether they hold for linux too (i was unable to reproduce the problem there, although the original bug report originates from linux). Any comments, and/or help are appreciated :) markus [1] http://bugs.gentoo.org/show_bug.cgi?id=352254 > > Some fixes have been made to GNU tar recently in that area, > and it's possible the bug has already been fixed. If you > have the time and inclination, you can try building from > the git repository <http://savannah.gnu.org/git/?group=tar> > and see whether that fixes things. >
